Abstract: A system is provided for training an acoustic model for use in speech recognition. In particular, such a system may be used to perform training based on a spoken audio stream and a non-literal transcript of the spoken audio stream. Such a system may identify text in the non-literal transcript which represents concepts having multiple spoken forms. The system may attempt to identify the actual spoken form in the audio stream which produced the corresponding text in the non-literal transcript, and thereby produce a revised transcript which more accurately represents the spoken audio stream. The revised, and more accurate, transcript may be used to train the acoustic model using discriminative training techniques, thereby producing a better acoustic model than that which would be produced using conventional techniques, which perform training based directly on the original non-literal transcript.

Abstract: A speech recognition client sends a speech stream and control stream in parallel to a server-side speech recognizer over a network. The network may be an unreliable, low-latency network. The server-side speech recognizer recognizes the speech stream continuously. The speech recognition client receives recognition results from the server-side recognizer in response to requests from the client. The client may remotely reconfigure the state of the server-side recognizer during recognition.

Abstract: A computer-based system includes a computer-processable definition of a region in a data set. The system identifies a region of the data set based on the definition of the region. The system provides output to a user representing a question and the identified region of the data set. The system may also automatically generate an answer to the question based on the question and the data set, and provide output to the user representing the answer. The system may generate the answer based on a subset of the data set, and provide output to the user representing the subset of the data set. The user may provide feedback on the first answer to the system, which the system may use to improve subsequent answers to the same and other questions, and to disable the system's automatic question-answering function in response to disagreement between the user and the system.

Abstract: An automatic speech recognition system includes an audio capture component, a speech recognition processing component, and a result processing component which are distributed among two or more logical devices and/or two or more physical devices. In particular, the audio capture component may be located on a different logical device and/or physical device from the result processing component. For example, the audio capture component may be on a computer connected to a microphone into which a user speaks, while the result processing component may be on a terminal server which receives speech recognition results from a speech recognition processing server.

Abstract: A dataset is searched using inclusion set criteria to produce an inclusion set and exclusion set criteria to produce an exclusion set. A set of unique content elements is identified from the inclusion set and the exclusion set. Metrics are derived from the inclusion set, exclusion set, and set of unique content elements, such as a measure, for each unique content element, of the absolute value of the difference between the percentage of records in the inclusion set containing the unique content element and the percentage of records in the exclusion set containing the unique content element. The unique content element set may be sorted and displayed in decreasing order of the above-referenced absolute value. The content element set may be filtered. Individual content elements may be excluded from the content set. A predictive model may be generated based on the resulting version of the content element set.

Abstract: Inputs provided into user interface elements of an application are observed. Records are made of the inputs and the state(s) the application was in while the inputs were provided. For each state, a corresponding language model is trained based on the input(s) provided to the application while the application was in that state. When the application is next observed to be in a previously-observed state, a language model associated with the application's current state is applied to recognize speech input provided by a user and thereby to generate speech recognition output that is provided to the application. An application's state at a particular time may include the user interface element(s) that are displayed and/or in focus at that time, and is determined by an operating system hooking component embedded in the automatic speech recognition system.

Abstract: Methods and systems are disclosed for tracking quality measures in abstracted documents. Embodiments include, determining, based on the abstracted content, a quality measure category, obtaining a quality measure definition corresponding to a quality measure included in the determined quality measure category, determining, based on keywords corresponding to the criterion, whether a portion of the abstracted content satisfies a criterion, recording, in association with the criterion, a reference to the portion of the abstracted content that satisfies the criterion; and selectively generating, a report including the query corresponding to the criterion, a query response, and the portion of the abstracted content satisfying the criterion.

Abstract: Techniques are disclosed for facilitating the process of proofreading draft transcripts of spoken audio streams. In general, proofreading of a draft transcript is facilitated by playing back the corresponding spoken audio stream with an emphasis on those regions in the audio stream that are highly relevant or likely to have been transcribed incorrectly. Regions may be emphasized by, for example, playing them back more slowly than regions that are of low relevance and likely to have been transcribed correctly. Emphasizing those regions of the audio stream that are most important to transcribe correctly and those regions that are most likely to have been transcribed incorrectly increases the likelihood that the proofreader will accurately correct any errors in those regions, thereby improving the overall accuracy of the transcript.

Abstract: An automatic speech recognition system includes an audio capture component, a speech recognition processing component, and a result processing component which are distributed among two or more logical devices and/or two or more physical devices. In particular, the audio capture component may be located on a different logical device and/or physical device from the result processing component. For example, the audio capture component may be on a computer connected to a microphone into which a user speaks, while the result processing component may be on a terminal server which receives speech recognition results from a speech recognition processing server.

Abstract: A computer system increases the efficiency with which billing codes may be generated based on a chart, such as a medical chart. The computer system provides the chart to a computer-assisted coding (CAC) module, which produces an initial set of billing codes and an initial assessment of the accuracy and/or completeness of the codes. The computer system decides whether to send the initial set of billing codes to an initial human reviewer. If the computer system sends the initial set of billing codes to the initial human reviewer, then the initial human reviewer reviews the chart and the output of the CAC module, and attempts to fix errors in the CAC output. The system provides the chart and the current (initial or modified) codes to a final human reviewer, who may be more highly skilled than the initial human reviewer, for final verification and modification.

Abstract: A computer-based system includes a computer-processable definition of a region in a data set. The system identifies a region of the data set based on the definition of the region. The system provides output to a user representing a question and the identified region of the data set. The system may also automatically generate an answer to the question based on the question and the data set, and provide output to the user representing the answer. The system may generate the answer based on a subset of the data set, and provide output to the user representing the subset of the data set. The user may provide feedback on the first answer to the system, which the system may use to improve subsequent answers to the same and other questions, and to disable the system's automatic question-answering function in response to disagreement between the user and the system.

Abstract: A system is provided for training an acoustic model for use in speech recognition. In particular, such a system may be used to perform training based on a spoken audio stream and a non-literal transcript of the spoken audio stream. Such a system may identify text in the non-literal transcript which represents concepts having multiple spoken forms. The system may attempt to identify the actual spoken form in the audio stream which produced the corresponding text in the non-literal transcript, and thereby produce a revised transcript which more accurately represents the spoken audio stream. The revised, and more accurate, transcript may be used to train the acoustic model, thereby producing a better acoustic model than that which would be produced using conventional techniques, which perform training based directly on the original non-literal transcript.

Abstract: An automatic speech recognizer is used to produce a structured document representing the contents of human speech. A best practice is applied to the structured document to produce a conclusion, such as a conclusion that required information is missing from the structured document. Content is inserted into the structured document based on the conclusion, thereby producing a modified document. The inserted content may be obtained by prompting a human user for the content and receiving input representing the content from the human user.

Abstract: A dataset is searched using inclusion set criteria to produce an inclusion set and exclusion set criteria to produce an exclusion set. A set of unique content elements is identified from the inclusion set and the exclusion set. Metrics are derived from the inclusion set, exclusion set, and set of unique content elements, such as a measure, for each unique content element, of the absolute value of the difference between the percentage of records in the inclusion set containing the unique content element and the percentage of records in the exclusion set containing the unique content element. The unique content element set may be sorted and displayed in decreasing order of the above-referenced absolute value. The content element set may be filtered. Individual content elements may be excluded from the content set. A predictive model may be generated based on the resulting version of the content element set.

Abstract: An automatic speech recognition system includes an audio capture component, a speech recognition processing component, and a result processing component which are distributed among two or more logical devices and/or two or more physical devices. In particular, the audio capture component may be located on a different logical device and/or physical device from the result processing component. For example, the audio capture component may be on a computer connected to a microphone into which a user speaks, while the result processing component may be on a terminal server which receives speech recognition results from a speech recognition processing server.

Abstract: A system is provided for training an acoustic model for use in speech recognition. In particular, such a system may be used to perform training based on a spoken audio stream and a non-literal transcript of the spoken audio stream. Such a system may identify text in the non-literal transcript which represents concepts having multiple spoken forms. The system may attempt to identify the actual spoken form in the audio stream which produced the corresponding text in the non-literal transcript, and thereby produce a revised transcript which more accurately represents the spoken audio stream. The revised, and more accurate, transcript may be used to train the acoustic model using discriminative training techniques, thereby producing a better acoustic model than that which would be produced using conventional techniques, which perform training based directly on the original non-literal transcript.

Abstract: A system is provided for training an acoustic model for use in speech recognition. In particular, such a system may be used to perform training based on a spoken audio stream and a non-literal transcript of the spoken audio stream. Such a system may identify text in the non-literal transcript which represents concepts having multiple spoken forms. The system may attempt to identify the actual spoken form in the audio stream which produced the corresponding text in the non-literal transcript, and thereby produce a revised transcript which more accurately represents the spoken audio stream. The revised, and more accurate, transcript may be used to train the acoustic model, thereby producing a better acoustic model than that which would be produced using conventional techniques, which perform training based directly on the original non-literal transcript.

Abstract: An automatic speech recognizer is used to produce a structured document representing the contents of human speech. A best practice is applied to the structured document to produce a conclusion, such as a conclusion that required information is missing from the structured document. Content is inserted into the structured document based on the conclusion, thereby producing a modified document. The inserted content may be obtained by prompting a human user for the content and receiving input representing the content from the human user.

Abstract: A dataset is searched using inclusion set criteria to produce an inclusion set and exclusion set criteria to produce an exclusion set. A set of unique content elements is identified from the inclusion set and the exclusion set. Metrics are derived from the inclusion set, exclusion set, and set of unique content elements, such as a measure, for each unique content element, of the absolute value of the difference between the percentage of records in the inclusion set containing the unique content element and the percentage of records in the exclusion set containing the unique content element. The unique content element set may be sorted and displayed in decreasing order of the above-referenced absolute value. The content element set may be filtered. Individual content elements may be excluded from the content set. A predictive model may be generated based on the resulting version of the content element set.

Abstract: Techniques are disclosed for facilitating the process of proofreading draft transcripts of spoken audio streams. In general, proofreading of a draft transcript is facilitated by playing back the corresponding spoken audio stream with an emphasis on those regions in the audio stream that are highly relevant or likely to have been transcribed incorrectly. Regions may be emphasized by, for example, playing them back more slowly than regions that are of low relevance and likely to have been transcribed correctly. Emphasizing those regions of the audio stream that are most important to transcribe correctly and those regions that are most likely to have been transcribed incorrectly increases the likelihood that the proofreader will accurately correct any errors in those regions, thereby improving the overall accuracy of the transcript.